Natural Vision Research Articles

Artificial intelligence: Friend or foe?

Artificial intelligence: Friend or foe?

Transformer for Skeleton-based action recognition: A review of recent advances

Skeleton-based action recognition has rapidly become one of the most popular and essential research topics in computer vision. The task is to analyze the characteristics of human joints and accurately classify their behaviors through deep learning technology. Skeleton provides numerous unique advantages over other data modalities, such as robustness, compactness, noise immunity, etc. In particular, the skeleton modality is extremely lightweight, which is especially beneficial for deep learning research in low-resource environments. Due to the non-European nature of skeleton data, Graph Convolution Network (GCN) has become mainstream in the past few years, leveraging the benefits of processing topological information. However, with the explosive development of transformer methods in natural language processing and computer vision, many works have applied transformer into the field of skeleton action recognition, breaking the accuracy monopoly of GCN. Therefore, we conduct a survey using transformer method for skeleton-based action recognition, forming of a taxonomy on existing works. This paper gives a comprehensive overview of the recent transformer techniques for skeleton action recognition, proposes a taxonomy of transformer-style techniques for action recognition, conducts a detailed study on benchmark datasets, compares the algorithm accuracy of standard methods, and finally discusses the future research directions and trends. To the best of our knowledge, this study is the first to describe skeleton-based action recognition techniques in the style of transformers and to suggest novel recognition taxonomies in a review. We are confident that Transformer-based action recognition technology will become mainstream in the near future, so this survey aims to help researchers systematically learn core tasks, select appropriate datasets, understand current challenges, and select promising future directions.

Supporting Humanitarian Crisis Decision Making with Reliable Intelligence Derived from Social Media Using AI

Abstract Recent advances in the field of artificial intelligence bring promising new capabilities that can substantially improve our ability to manage complex and evolving situations in the face of uncertainty. Humanitarian crises exemplify such situations, and the pervasiveness of social media renders it one of the most abundant sources of real-time information available. However, it is quite a difficult task to condense a body of social media posts into useful information quickly. In this paper we consider the challenge of using social media reports to provide a reliable, real-time situational awareness in the management of humanitarian crises. Effectively addressing this challenge requires extracting only the relevant information out of text and images in individual social media posts, fusing this information together into actionable information points for decision makers, and providing an assessment of the trustworthiness of this information. We propose a general solution framework and discuss a system developed in collaboration with NATO which combines state-of-the-art deep learning, natural language processing, computer vision, and information fusion models to provide a reliable, actionable, real-time situational awareness for supporting decision making in humanitarian crisis logistics. In addition to the technical approach, we also discuss important practical aspects of this project including the development and validation process, challenges encountered along the way, and key lessons learned.

Distinct dorsal and ventral streams for binocular rivalry dominance and suppression revealed by magnetoencephalography.

Binocular rivalry is an example of bistable visual perception extensively examined in neuroimaging. Magnetoencephalography can track brain responses to phasic visual stimulations of predetermined frequency and phase to advance our understanding of perceptual dominance and suppression in binocular rivalry. We used left and right eye stimuli that flickered at two tagging frequencies to track their respective oscillatory cortical evoked responses. We computed time-resolved measures of coherence to track brain responses phase locked with stimulus frequencies and with respect to the participants' indications of alternations of visual rivalry they experienced. We compared the brain maps obtained to those from a non-rivalrous control replay condition that used physically changing stimuli to mimic rivalry. We found stronger coherence within a posterior cortical network of visual areas during rivalry dominance compared with rivalry suppression and replay control. This network extended beyond the primary visual cortex to several retinotopic visual areas. Moreover, network coherence with dominant percepts in primary visual cortex peaked at least 50 ms prior to the suppressed percept nadir, consistent with the escape theory of alternations. Individual alternation rates were correlated with the rate of change in dominant evoked peaks, but not for the slope of response to suppressed percepts. Effective connectivity measures revealed that dominant (respectively, suppressed) percepts were expressed in dorsal (respectively ventral) streams. We thus demonstrate that binocular rivalry dominance and suppression engage distinct mechanisms and brain networks. These findings advance neural models of rivalry and may relate to more general aspects of selection and suppression in natural vision.

A transformer architecture for retention time prediction in liquid chromatography mass spectrometry-based proteomics.

Accurate retention time (RT) prediction is important for spectral library-based analysis in data-independent acquisition mass spectrometry-based proteomics. The deep learning approach has demonstrated superior performance over traditional machine learning methods for this purpose. The transformer architecture is a recent development in deep learning that delivers state-of-the-art performance in many fields such as natural language processing, computer vision, and biology. We assess the performance of the transformer architecture for RT prediction using datasets from five deep learning models Prosit, DeepDIA, AutoRT, DeepPhospho, and AlphaPeptDeep. The experimental results on holdout datasets and independent datasets exhibit state-of-the-art performance of the transformer architecture. The software and evaluation datasets are publicly available for future development in the field.

Artificial Intelligent Techniques for Identifying the Cause of Disturbances in the Power Grid

The intricacy of the power system configuration, coupled with the contemporary trends in power generation and demand, renders the attainment of adequate supply quality a daunting task for distribution companies. Several regulations govern the power quality (PQ) of the electrical system. In addition, EN-50160 outlines the voltage characteristics supplied by public electricity networks, which directly impact distribution companies. Secondly, the EN-61000 standards series regulates the electromagnetic compatibility of network-connected devices, which affects the loads. Both power companies and device manufacturers are responsible for ensuring and being impacted by the quality of supply. Artificial Intelligence (AI) techniques refer to a variety of methods and algorithms that enable machines to perform tasks that typically require human-like intelligence, such as perception, reasoning, learning, and decision-making. AI techniques include machine learning, natural language processing, computer vision, robotics, expert systems, and other approaches that use algorithms to analyze and understand complex data, recognize patterns, and make predictions or decisions based on that data Notwithstanding the regulations, there are still unresolved aspects of the supply quality, one of the most significant being the location of the origin of disturbances. This article presents an investigation of the main techniques used to identify the cause of disturbances and locate their origin in the power grid.

People vs. peoples: sacrifice and the foundations for sovereignty in 1640’s England and contemporary northeast Syria

ABSTRACT Engaging with state-centric perspectives on sovereignty, and particularly Giorgio Agamben's work, this article argues that despite sovereignty's permanence at the level of the state, the principles upon which it is built can change dramatically. Further, such change may in turn greatly transform the reach and power of the state itself. Using the trial of King Charles in 1649 as a case for of how divine right was replaced by embryonic popular sovereignty through an act of sacrifice, the article contends that several new revolutionary movements are now in turn aiming to overturn popular sovereignty in a similar way. Drawing on secondary material and fieldwork contextualization, the article contends that the Kurdish movement in Syria is currently struggling to disaggregate the state's people into a host of peoples, whose opportunities for political participation depend upon conforming to an ideologically construed vision of human nature. This, the article suggests, may set the frame for revolutionary resistance in the future.

Red Alarm for Pre-trained Models: Universal Vulnerability to Neuron-level Backdoor Attacks

The pre-training-then-fine-tuning paradigm has been widely used in deep learning. Due to the huge computation cost for pre-training, practitioners usually download pre-trained models from the Internet and fine-tune them on downstream datasets, while the downloaded models may suffer backdoor attacks. Different from previous attacks aiming at a target task, we show that a backdoored pre-trained model can behave maliciously in various downstream tasks without foreknowing task information. Attackers can restrict the output representations (the values of output neurons) of trigger-embedded samples to arbitrary predefined values through additional training, namely neuron-level backdoor attack (NeuBA). Since fine-tuning has little effect on model parameters, the fine-tuned model will retain the backdoor functionality and predict a specific label for the samples embedded with the same trigger. To provoke multiple labels in a specific task, attackers can introduce several triggers with predefined contrastive values. In the experiments of both natural language processing (NLP) and computer vision (CV), we show that NeuBA can well control the predictions for trigger-embedded instances with different trigger designs. Our findings sound a red alarm for the wide use of pre-trained models. Finally, we apply several defense methods to NeuBA and find that model pruning is a promising technique to resist NeuBA by omitting backdoored neurons.

Maternity and Motherlessness

Maternity and Motherlessness Anna McFarlane (bio) Wendy C. Nielsen. Motherless Creations: Fictions of Artificial Life, 1650–1890. Routledge, 2022. 262 pp, 13 b/w illus. $170 hc, $48.95 ebk. Renae L. Mitchell. Maternity in the Post-Apocalypse: Novelistic Re-Visions of Dystopian Motherhood. Lexington, 2022. 162 pp. $95 hc, $45 ebk. There has been growing interest in the figure of the child in recent engagements with sf and the possibility of a future, especially in the age of the climate crisis and particularly influenced by Lee Edelman’s polemical contribution to queer theory, No Future: Queer Theory and the Death Drive (2004). Rebekah Sheldon’s The Child to Come: Life After the Human Catastrophe (2016), for example, works from Edelman’s arguments to claim that the figure of the child is used to shut down the possibilities for political change in conservative ways that amount to an emergency in the face of a climate crisis that remains unaddressed by our contemporary political inertia. Sheldon mobilizes science fiction to make her point, as does Heather Latimer, whose 2021 reading of the film Arrival (Villeneuve 2016) as a text that queers pregnancy and maternity makes a valuable contribution to these discussions. Maggie Nelson’s queering of her own pregnancy in her memoir The Argonauts (2015) also contributes to contemporary discussions about maternity and how its heteronormative and restrictive connotations might be read and overturned in literature and in public life. Against this backdrop, it is unsurprising that the figure of the mother and the life stages of pregnancy and maternity continue to be analyzed from new perspectives in sf studies. Two new monographs—Wendy C. Nielsen’s Motherless Creations and Renae L. Mitchell’s Maternity in the Post-Apocalypse—tackle these issues from the perspectives of posthumanism and psychoanalysis respectively. Nielsen’s book starts with a story and a question. The story goes that René Descartes once set sail accompanied by a figure whom he claimed to be his daughter Francine, but who looked wrong in the light of day; this caused fellow passengers to cross themselves and blame the poor weather on the supposed automaton until it was tossed overboard, to follow in the ship’s wake like a bad omen. Nielsen tells us that the story was a fabrication (1), but this did not stand in the way of its longevity: the desire to frame Descartes as the father of artificial intelligence meant that the tale was repeated, and the truth—that the story had been concocted to hide the fact that Descartes fathered an illegitimate child—was lost in the process. Nielsen sees this story as emblematic of a common narrative of artificial intelligence as the being that [End Page 105] is produced by a man, thereby eliding the role of the mother and the role of women more generally. She asks, “Why does early speculative fiction eliminate women’s roles as mothers?” (1), and her book is a series of examples and attempts at explanation. She also asks, “In what ways do beings created without mothers sustain or challenge traditional concepts of gendered or racial identity; what it means to be a mother, father, or creator, and the nature of birth, imagination, and creation?” (1–2). Nielsen takes her examples from English, American, French, and German literature. She differentiates the (potentially) bodiless programs of Artificial Intelligence from the embodied, anthropomorphic Artificial Life—or ALife—that “has emerged as a discipline in robotics and as an area of critical inquiry of the posthuman in philosophy, literature, and gender studies” (2; emphasis in original). Nielsen’s rationale for the historical sweep of her subject is that she dates 1650 as a time when automata, and therefore ALife, became the subject of science rather than magic, as androids with the potential ability to mimic humans entered the market for the first time (2). Nielsen argues that “in Western literature before 1890, the recurring figure of the motherless creation represents a desire to create the perfect child and sustains pseudoscientific beliefs about the birthing body” (3)—for example, that a man’s production of new life will be a safer and more rational process than the chaotic danger of old-fashioned...

Memristor-Based Light-Weight Transformer Circuit Implementation for Speech Recognizing

Transformer network (TN) is a promising model widely used for natural language processing (NLP), computer vision (CV), and audio processing (AP). However, the large number of multiples and additions (multi-adds) in TN leads to a massive number of parameters in the TN model, causing an increasing number of hardware resource overhead and thus making it difficult to deploy the TN model to edge devices. To alleviate this problem, this paper proposes a circuit implementation scheme for a memristor-based lightweight transformer network (MLTN). In addition, the speech recognition task based on the speech command dataset is implemented through the modular hardware design approach, which provides an effective lightweight hardware implementation scheme for the edge devices. Further, the retention, noise immunity, variability, and stability of the designed circuit are analyzed by simulation experiments with LTSPICE. The results show that the proposed MLTN circuit not only has a lighter network model but also has significant recognition accuracy with different hardware scales. Specifically, MLTN-40 can achieve 96.18% simulation accuracy, while the energy overhead is only 3.8% of the ARM framework, which is commonly used for edge devices, and 39.9% of the traditional server (GPU), making it a promising solution for implementing speech recognition on edge devices.

Advancing Federated Learning with Granular Computing

Over the recent years, we have been witnessing spectacular achievements of Artificial Intelligence (AI) and Machine Learning (ML), in particular. We have seen highly visible accomplishments encountered in natural language processing and computer vision impacting numerous areas of human endeavours. Being driven inherently by the technologically advanced learning and architectural developments, ML constructs are highly impactful coming with far reaching consequences; just to mention autonomous vehicles, health care imaging, decision-making processes in critical areas, among others. The quality of ML architectures and credibility of generated results are inherently implied by the nature, quality, and amount of available data. The credibility of ML models and confidence quantified their results are also of paramount concern to any critical application. In this study, we advocate that the credibility (confidence) of results produced by ML constructs is inherently expressed in the form of information granules. Several development scenarios are carefully revisited including those involving constructs in statistics (confidence and prediction intervals), probability (Gaussian process models), and granular parameters (fuzzy sets and interval techniques). We augment the commonly encountered and challenging category of applications of ML referred to as federated learning where the aspect of quality of the model and its results calls for a thorough assessment.

Guerre by Louis-Ferdinand Céline

Reviewed by: Guerre by Louis-Ferdinand Céline Jane E. Evans CÉline, Louis-Ferdinand. Guerre. Gallimard, 2022. ISBN 978-2-07-298322-1. Pp. 191. Literary scholars contend that the term "autofiction" better captures the nature of life-writing than the term "autobiography," with the former allowing for the enhancement and downplaying of certain events by the writer for a variety of reasons, modesty, lost memories, and creativity, among them. Louis-Ferdinand Céline's novel, Guerre, exemplifies autofiction as it interweaves fact with fantasy. Based on his 1914 experiences in the Great War, Céline's first-person account was composed twenty years later. The narrator, "Louis-Ferdinand," "Ferdinand," or "Loulou," compelled to share his war experiences, declares early in the story, "J'ai attrapé la guerre dans ma tête. Elle est enfermée dans ma tête" (26). His words convey the effects of having been bombed, both literally and figuratively: his head reverberates from the explosions and the deafness in his left ear; he also expresses the idea that sharing his experiences will be psychologically beneficial. While he leads people to believe that he was shot in the head as well as in the arm, his concussion resulted from falling against a tree during the bombardment. If he later adds, "Mais faut que je raconte tout" (34), he refrains from promising accuracy. Rather, he addresses the reader as follows, "J'ai appris à faire de la musique […] et, vous le voyez, de la belle littérature aussi, avec des petits morceaux d'horreur arrachés au bruit qui n'en finira jamais" (28). Thus, events documented by letters from the French Army to Louis-Ferdinand's father regarding his son's "héroïsme" (97) towards his fellow soldiers and grave injury to his right arm appear alongside romanticized elements, such as the eagerness of nurse L'Espinasse to gauge the wounded soldiers' potential for recovery according to how well they respond to her manual stimulation of their genitals. In real life, Loulou's nurse, Alice David, struck up a friendship with him and behaved like a big sister (14). The narrator's receipt of a military medal of valor, another verifiable occurrence, contrasts sharply with the fantasized verbal altercations between Céline's wardmate, Bébert, and his prostitute wife, Angèle, that purportedly end their marriage. Additional instances of diverting the truth include modifications of real names in Part Two of his récit: "Bébert" becomes "Cascade" while the narrator's mother, "Célestine" in Part One, is renamed "Clémence" in the subsequent section. These changes attest to creativity on the narrator's part, although he does not explain why. He does, however, caution the reader about the fallibility of memory with the passage of time: "À tant d'années passées le souvenir des choses, bien précisément, c'est un effort. Ce que les gens ont dit c'est presque tourné des mensonges. Faut se méfier. C'est putain le passé, ça fond dans la rêvasserie" (117). This avowal underscores the imaginary nature of his characters' dialogues once again. At the end of Guerre, the narrator claims to have left the French hospital in Peurdu-sur-la-Lys to continue his convalescence in London directly, although the truth is that Louis-Ferdinand spent several months in a Parisian hospital before making his way to England (169). [End Page 224] Jane E. Evans University of Texas, El Paso Copyright © 2023 American Association of Teachers of French

Imagination as a Facet of Openness/Intellect: A New Scale Differentiating Experiential Simulation and Conceptual Innovation

ABSTRACT Previous research has investigated the nature of imagination as a construct related to multiple forms of higher-order cognition. Despite the emergence of various conceptualizations of imagination, few attempts have been made to explore the structure of imagination as a trait in the context of existing hierarchically-nested personality dimensions. We present the Dual Facet Imagination Scale (DFIS) for measuring trait imagination that distinguishes between experiential simulation and conceptual innovation, aligned with the two major subfactors (aspects) of the Big Five dimension Openness/Intellect. In two samples (N = 210 & 5771), we provide evidence of a consistent factor structure distinguishing experiential and conceptual descriptions of imagination, as well as validity as measures of facets of Openness and Intellect. The DFIS provides a trait measure of two major forms of imagination in line with mainstream models of the hierarchical structure of personality.

Beyond Maṣlaḥah

This paper focuses on maṣlaḥah (benefit or well-being) and adab (righteous behavior or character) as ethically intertwined concepts that are discussed by classical Muslim scholars in relation to the acquisition of wealth (kasb) and overall economic engagement. Particularly in certain works of al-Shaybāni (d. 805), al-Muḥāsibī (d. 857), Ibn Abī al-Dunyā (d. 894), al-Māwardī (d. 1058), and al-Ghazālī (d. 1111), personal piety is closely related to righteous economic behavior under the banner of adab’s moral stipulations. In light of al-Ghazālī’s understanding of economic provision as part of his overall theory of eternal happiness (saʿādah), the concept of maṣlaḥah can be analyzed in the context of adab as an extension of Sharīʿa law. While maṣlaḥah is from a legal standpoint crucial for safeguarding economic activities and preserving wealth, concomitantly, in this paper I treat maṣlaḥah as a derivative of adab and its holistic vision of human nature. In particular, I address what constitutes economic provision as an ethical endeavor in selected classical texts; and how the concept of adab preserves and enhances economic behavior as conceived by classical Muslim scholars.

Attention-Based Adversarial Robust Distillation in Radio Signal Classifications for Low-Power IoT Devices

Due to great success of transformers in many applications, such as natural language processing and computer vision, transformers have been successfully applied in automatic modulation classification. We have shown that transformer-based radio signal classification is vulnerable to imperceptible and carefully crafted attacks called adversarial examples. Therefore, we propose a defense system against adversarial examples in transformer-based modulation classifications. Considering the need for computationally efficient architecture particularly for Internet of Things (IoT)-based applications or operation of devices in an environment where power supply is limited, we propose a compact transformer for modulation classification. The advantages of robust training such as adversarial training in transformers may not be attainable in compact transformers. By demonstrating this, we propose a novel compact transformer that can enhance robustness in the presence of adversarial attacks. The new method is aimed at transferring the adversarial attention map from the robustly trained large transformer to a compact transformer. The proposed method outperforms the state-of-the-art techniques for the considered white-box scenarios, including the fast gradient method and projected gradient descent attacks. We have provided reasoning of the underlying working mechanisms and investigated the transferability of the adversarial examples between different architectures. The proposed method has the potential to protect the transformer from the transferability of adversarial examples.

Forgotten whales, fading codfish: Perceptions of ‘natural’ ecosystems inform visions of future recovery

Perceptions of past ecological change affect views of current ecosystem state, but how do baselines help to shape stakeholders' visions of an idealized future? Here, we investigate links between perceptions of natural baselines and visions for the nearshore Gulf of Maine among a key stakeholder group, active lobster fishers. We ask three related questions: (1) What do fishers perceive as a ‘natural’ Gulf of Maine? (2) How do perceptions of the past predict individual and collective visions of an ideal future? and (3) How is existing management perceived as supporting these visions? We found that fishers perceived the ecosystem to be ‘natural’ an average of one decade before they started fishing. Three species dominated views of natural systems: cod Gadus morhua, lobster Homarus americanus, and herring Clupea harengus, but while long-time fishers associated abundant cod with a natural nearshore Gulf of Maine, memories of a historically cod-rich Gulf of Maine were fading among some younger fishers who began their careers after the cod crash in the 1990s. Perceptions of ‘natural’ ecosystems dictated future visions for the majority of taxa; on average, fishers remembered and desired abundant cod and herring, but perceived halibut Hippoglossus hippoglossus and endangered right whales Eubalaena glacialis to have always been rare. Fishers described a vision for the future based on views of past ecological and social baselines, including fisheries deconsolidation and diversification, but expressed a lack of shared vision with and trust in federal management institutions to achieve these goals. In particular, memories of cod abundance in the 1970s and 1980s were coupled with memories of a diversified and accessible fishery, but fishers doubted that the recovery of cod would result in their restored access to cod fisheries. Together our results demonstrate that past personal experiences limit perceptions of what is possible, highlighting both the value and limitations of local ecological knowledge in places that have experienced ecological change over centuries. They also demonstrate how stakeholder perceptions of both social and ecological baselines shape visions for future ecosystems but are mediated by contemporary issues like trust in institutions and fisheries access. Read the free Plain Language Summary for this article on the Journal blog.

A Survey on RISC-V-Based Machine Learning Ecosystem

In recent years, the advancements in specialized hardware architectures have supported the industry and the research community to address the computation power needed for more enhanced and compute intensive artificial intelligence (AI) algorithms and applications that have already reached a substantial growth, such as in natural language processing (NLP) and computer vision (CV). The developments of open-source hardware (OSH) and the contribution towards the creation of hardware-based accelerators with implication mainly in machine learning (ML), has also been significant. In particular, the reduced instruction-set computer-five (RISC-V) open standard architecture has been widely adopted by a community of researchers and commercial users, worldwide, in numerous openly available implementations. The selection through a plethora of RISC-V processor cores and the mix of architectures and configurations combined with the proliferation of ML software frameworks for ML workloads, is not trivial. In order to facilitate this process, this paper presents a survey focused on the assessment of the ecosystem that entails RISC-V based hardware for creating a classification of system-on-chip (SoC) and CPU cores, along with an inclusive arrangement of the latest released frameworks that have supported open hardware integration for ML applications. Moreover, part of this work is devoted to the challenges that are concerned, such as power efficiency and reliability, when designing and building application with OSH in the AI/ML domain. This study presents a quantitative taxonomy of RISC-V SoC and reveals the opportunities in future research in machine learning with RISC-V open-source hardware architectures.

Spatiochromatic and temporal natural image statistics modelling: Applications from display analysis to neural networks

Natural image statistics are well known to have a spatial frequency power spectra that has a 1/f^a behavior, with a typically stated as between 2 and 4. This indicates an invariance to scale. Further work has theorized how the visual system is tuned for such statistics in visual cortex (V1) [1]. Color image statistics also show an invariance to scale [2]. The luminance histogram is typically understood to be log normal with respect to luminance, although for HDR images, a subcomponent with skew toward much higher luminances is observed. Color statistics were initially described at the simplest level via the gray world hypothesis [3], but more details are now available, even at the hyperspectral [4]. The a power function for HDR was found to increase from the lower values of 2 to more typical values of 4 and 5 [5]. For temporal statistics, the data tends to be measured primarily for media, with a 1/f^a for scene cut statistics [6], and temporal frequency and temporal frequency for media with a focus on the motion statistics via optical flow [7]. Statistics for purely natural as well as human made environments (e.g., buildings and the resulting perspective geometry) have been studied, each having different orientation statistics [8]. The use of image statistics for standardized assessment of television power consumption was used to replace test targets, which were often detected and used to lower TV power consumption in well known cheating schemes. To prevent this, a short test video that had luminance statistics matching 48 hours of broadcast content was generated and used for TV power testing [9]. The highly adaptative nature of current TVs (power limiting, dual modulation, dynamic response) has motivated researchers to incorporate complex noise fields following natural image statistics into measurement targets [10,11]. One particular natural image statistic-based still image test target (dead leaves) is widely used in camera optics and sensor development. Algorithm development and testing for image and video processing has almost always been ad hoc, with a mixture of geometric test targets and hand selected test images, sometimes aiming to be corner cases, sometimes not. More recently, large data sets of images have been used to train various neural network models for tasks such as super resolution, bit rate compression, and dynamic range mapping. However, images are not ergodic, and possibly not even wide-sense stationary. We propose the use of imagery based on noise following the natural image statistics for spatio-chromatic (and temporal) to compactly probe the wide variety of image possibilities for algorithmic development, in addition to the existing uses for image capture and display analysis. While we don’t suggest replacing actual practical imagery, we believe such noise fields can augment image algorithm analysis. To address the problem of non-ergodicity, we allow the basic power term a in the natural image statistic model to vary over a large range in a video, such that it includes the extremes of white noise and low frequency gradients. We use color image statistic models that include decorrelated colors to generate the RGB video. We will present results for traditional adaptive data compression (with chromatic subsampling), as well as a more contemporary neural network approach (Neural Fields [12]) as applied to upscaling and denoising. We analyze the results both visually and through several recent color image quality models. Field DJ. Relations between the statistics of natural images and the response properties of cortical cells. J. Opt. Soc. Am. A, 1987; 4:2379-2394 C. Parraga, T. Troscianko, and D.J. Tolhurst (2002) spatiochromatic properties of natural images and human vision. Current Biology V 12 R. M. Evans, Method for correcting photographic color prints, US Patent 2,571,697 (1951) A. Chakrabarti and T. Zickler (2011) Statistics of real-world hyperspectral images CVPR R. Dror, A. Willsky, and E. Adelson (2004) statistical characterization of real-world illumination. JOV V4 J. Cutting (2019) Sequences in popular cinema generate inconsistent event segmentation. Attn. Percept. And Psycho. V 81. D. Lee, H. Ko, J. Kim, and A. Bovik (2021) On the space-time statistics of motion pictures. JOSA A V 38 #7 A. Torralba and A. Oliva (2003) Statistics of natural image categories, Network: Computational Neural Systems 14 391-412 International Electrotechnical Commission, IEC 62087:2008(E), “Methods of measurement for the power consumption of audio, video, and related Equipment. Kunkel T, Daly S. 57-1: Spatiotemporal Noise Targets Inspired by Natural Imagery Statistics. SID Symposium Digest of Technical Papers, 2020, 51:842-845. Kunkel, T, Friedrich, F. Utilizing advanced spatio-temporal backgrounds with dynamic test signals for high dynamic range display metrology. J Soc Inf Display. 2022; 30( 5): 423– 432. https://doi.org/10.1002/jsid.1125 Yiheng Xie1, Towaki Takikawa, Shunsuke Saito, Or Litany, Shiqin Yan, Numair Khan, Federico Tombari, James Tompkin, Vincent Sitzmann, Srinath Sridhar1, "Neural Fields in Visual Computing and Beyond", Eurographics / CGF State-of-the-Art Report, 2022.

Review on Machine Learning Strategies for Real-World Engineering Applications

As we enter the Industry 5.0 era, enormous volumes of data are being created across digital systems. Machine learning techniques have recently achieved immense success in areas such as intelligent control, decision-making, speech recognition, natural language processing, computer graphics, and computer vision. This despite the significant challenge of analyzing and interpreting massive datasets. Owing to their strong performance, deep learning and machine learning algorithms have become widely deployed across various real-time engineering applications. Developing working knowledge of machine learning is now critical for building automated, smart systems that can process data in domains like healthcare, cybersecurity, and intelligent transportation. There exist multiple strategies in machine learning, including reinforcement learning, semi-supervised learning, unsupervised learning, and supervised learning algorithms. This research provides a comprehensive examination of leveraging machine learning for managing real-time engineering systems, with the goal of augmenting their capabilities and intelligence. It contributes to the understanding of how different machine learning approaches can be applied in real-world use cases like cybersecurity, healthcare, and intelligent transportation. Additionally, it highlights ongoing research objectives and difficulties that machine learning techniques encounter while tackling real-world systems. This research serves both industry professionals and academics as a reference, while technically benchmarking decision-making across different application areas and real-world scenarios.

Deep learning based on self-supervised pre-training: Application on sandstone content prediction

Deep learning has been widely used in various fields and showed promise in recent years. Therefore, deep learning is the future trend to realize seismic data’s intelligent and automatic interpretation. However, traditional deep learning only uses labeled data to train the model, and thus, does not utilize a large amount of unlabeled data. Self-supervised learning, widely used in Natural Language Processing (NLP) and computer vision, is an effective method of learning information from unlabeled data. Thus, a pretext task is designed with reference to Masked Autoencoders (MAE) to realize self-supervised pre-training of unlabeled seismic data. After pre-training, we fine-tune the model to the downstream task. Experiments show that the model can effectively extract information from unlabeled data through the pretext task, and the pre-trained model has better performance in downstream tasks.